Defect information determination method and device of power transmission line, equipment, storage medium and program product

Abstract

The invention relates to a defect information determination method and device of a power transmission line, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps of obtaining a line image of a power transmission line and image information corresponding to the line image; the image information is used for indicating equipment in the power transmission line associated with the line image and attribute information of the equipment; performing image processing on the line image by using the image information to obtain a line image after image processing; inputting the line image after image processing into a defect identification model to obtain defect identification information of the power transmission line output by the defect identification model, and determining defect information of the power transmission line based on the defect identification information; the defect information is used for indicating at least one of a defect position, a defect type and a defect grade of the power transmission line. The method can improve the efficiency of determining the defect information of the power transmission line.

Description

Technical Field

[0001] This application relates to the field of power grid technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for determining defect information of transmission lines. Background Technology

[0002] As the core equipment for power transmission in the power system, transmission lines are widely distributed in complex outdoor environments. Their long-term stable operation directly determines the reliability and security of the power supply system. However, under the influence of multiple factors such as natural erosion, equipment aging, and external disturbances, transmission lines are prone to various defects such as broken conductor strands, damaged insulators, and corrosion of towers. Therefore, it is necessary to identify defects in transmission lines to determine their defect information.

[0003] In existing technologies, defect identification of large transmission lines mostly relies on manual inspections, and the defect information of the transmission lines is determined based on the inspection results. However, this method suffers from low efficiency. Summary of the Invention

[0004] Therefore, it is necessary to provide a more efficient method, apparatus, computer equipment, computer-readable storage medium, and computer program product for determining defect information of transmission lines, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for determining defect information of transmission lines, including:

[0006] Obtain the line image of the transmission line and the corresponding image information; the image information is used to indicate the equipment in the transmission line associated with the line image and the attribute information of the equipment;

[0007] Image processing is performed on the line image using image information to obtain the processed line image.

[0008] The processed line image is input into the defect identification model to obtain the defect identification information of the transmission line output by the defect identification model, and the defect information of the transmission line is determined based on the defect identification information; the defect information is used to indicate at least one of the defect location, defect type and defect level of the transmission line.

[0009] In one embodiment, the defect identification model includes a feature extraction unit, a candidate box generation unit, a confidence assessment unit, and a classification and grading unit. The process involves inputting a processed line image into the defect identification model to obtain defect identification information of the transmission line output by the model. This includes: inputting the processed line image into the feature extraction unit to obtain a multi-scale feature map output by the feature extraction unit; inputting the multi-scale feature map into the candidate box generation unit to obtain candidate boxes output by the candidate box generation unit; the candidate boxes are used to identify candidate defect regions in the multi-scale feature map; inputting the candidate boxes and the multi-scale feature map into the confidence assessment unit to obtain confidence information of the candidate defect regions output by the confidence assessment unit, and determining a target candidate defect region from the candidate defect regions based on the confidence information; and inputting the target candidate defect region into the classification and grading unit to obtain defect identification information of the transmission line output by the classification and grading unit.

[0010] In one embodiment, the feature extraction unit includes a backbone convolutional layer, a feature filtering activation layer, and a multi-scale feature fusion layer. The image-processed line image is input to the feature extraction unit to obtain a multi-scale feature map output by the feature extraction unit. This includes: inputting the image-processed line image into the backbone convolutional layer for convolution and downsampling to obtain an initial first-scale feature map, an initial second-scale feature map, and an initial third-scale feature map; inputting the initial first-scale feature map, the initial second-scale feature map, and the initial third-scale feature map into the corresponding feature filtering activation layer for feature filtering and enhancement to obtain a target first-scale feature map, a target second-scale feature map, and a target third-scale feature map; and inputting the target first-scale feature map, the target second-scale feature map, and the target third-scale feature map into the multi-scale feature fusion layer for upsampling and feature concatenation to obtain a multi-scale feature map.

[0011] In one embodiment, a multi-scale feature map is input to a candidate box generation unit to obtain candidate boxes output by the candidate box generation unit, including: acquiring defect size sample data of the transmission line, and performing cluster analysis on the defect size sample data based on a clustering algorithm to obtain a candidate box size set; and determining candidate boxes based on the candidate box size set and the multi-scale feature map.

[0012] In one embodiment, image processing is performed on a line image using image information to obtain a processed line image, including: determining the equipment type of the equipment in the associated transmission line based on the image information; and adjusting the resolution of the line image according to the equipment type to obtain the processed line image.

[0013] In one embodiment, determining the defect information of a transmission line based on defect identification information includes: determining whether the defect identification information meets preset verification conditions; if it does, determining the defect information of the transmission line based on the defect identification information; if it does not, correcting the defect identification information and determining the defect information of the transmission line based on the corrected defect identification information.

[0014] Secondly, this application also provides a device for determining defect information of transmission lines, comprising:

[0015] The acquisition module is used to acquire the line image of the transmission line and the corresponding image information; the image information is used to indicate the equipment in the transmission line associated with the line image and the attribute information of the equipment;

[0016] The execution module is used to perform image processing on the line image using image information to obtain the processed line image.

[0017] The determination module is used to input the image of the line after image processing into the defect recognition model to obtain the defect recognition information of the transmission line output by the defect recognition model, and to determine the defect information of the transmission line based on the defect recognition information; the defect information is used to indicate at least one of the defect location, defect type and defect level of the transmission line.

[0018] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the embodiments of the first aspect above.

[0019] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any of the embodiments of the first aspect above.

[0020] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the embodiments of the first aspect above.

[0021] The aforementioned method, apparatus, computer equipment, computer-readable storage medium, and computer program product for determining defect information of transmission lines first acquire a line image of the transmission line and corresponding image information. The image information indicates the equipment in the transmission line associated with the line image and the equipment's attribute information. Then, the image information is used to process the line image to obtain a processed line image. Next, the processed line image is input into a defect identification model to obtain the defect identification information of the transmission line output by the defect identification model. Based on the defect identification information, the defect information of the transmission line is determined. The defect information indicates at least one of the following: defect location, defect type, and defect level. The method for determining defect information of transmission lines provided in this application can determine the defect information of the transmission line based on the line image of the transmission line and the corresponding image information, effectively improving the efficiency of defect information determination compared to the prior art's method of determining defect information through manual inspection. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a method for determining defect information of a transmission line in one embodiment;

[0024] Figure 2 This is a flowchart illustrating a method for obtaining defect identification information in one embodiment;

[0025] Figure 3 This is a flowchart illustrating a method for obtaining multi-scale feature maps in one embodiment;

[0026] Figure 4 This is a flowchart illustrating a method for obtaining candidate boxes in one embodiment;

[0027] Figure 5 This is a flowchart illustrating a method for obtaining a processed line image in one embodiment.

[0028] Figure 6 This is a flowchart illustrating a method for determining defect information of a transmission line in one embodiment;

[0029] Figure 7 This is a flowchart illustrating a method for determining defect information of a transmission line in another embodiment;

[0030] Figure 8 This is a structural block diagram of a device for determining defect information of a transmission line in one embodiment;

[0031] Figure 9 This is an internal structural diagram of a computer device in one embodiment;

[0032] Figure 10 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0034] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0035] As the core equipment for power transmission in the power system, transmission lines are widely distributed in complex outdoor environments. Their long-term stable operation directly determines the reliability and security of the power supply system. However, under the influence of multiple factors such as natural erosion, equipment aging, and external disturbances, transmission lines are prone to various defects such as broken conductor strands, damaged insulators, and corrosion of towers. Therefore, it is necessary to identify defects in transmission lines to determine their defect information.

[0036] In existing technologies, defect identification of large transmission lines mostly relies on manual inspections, and the defect information of the transmission lines is determined based on the inspection results. However, this method suffers from low efficiency.

[0037] In view of this, this application provides a method for determining defect information of transmission lines. First, a line image of the transmission line and corresponding image information are acquired. The image information indicates the equipment in the transmission line associated with the line image and the equipment's attribute information. Then, the image information is used to process the line image to obtain a processed line image. Next, the processed line image is input into a defect recognition model to obtain the defect recognition information of the transmission line output by the defect recognition model. Based on the defect recognition information, the defect information of the transmission line is determined. The defect information indicates at least one of the following: defect location, defect type, and defect level. The method for determining defect information of transmission lines provided in this application can determine the defect information of the transmission line based on the line image and corresponding image information, effectively improving the efficiency of defect information determination compared to the prior art's method of determining defect information through manual inspection.

[0038] The method for determining defect information of transmission lines provided in this application can be executed by a computer device, which can be a terminal or a server.

[0039] In one exemplary embodiment, such as Figure 1 As shown, a method for determining defect information of transmission lines is provided, which includes the following steps:

[0040] Step 101: Obtain the line image of the transmission line and the corresponding image information.

[0041] Optionally, a line image of a transmission line refers to an image that includes the transmission line itself and its auxiliary equipment.

[0042] Image information can be used to indicate the equipment and equipment attributes in the transmission line associated with the line image.

[0043] For example, image information may include equipment type, associated line information, equipment attributes, and shooting assistance information. Equipment type may include, for example, insulator, 110kV conductor, or tension tower; associated line information may include, for example, line number "220kV Hongheng Line" and the branch office "XX Branch"; equipment attributes may include, for example, tower number "156", equipment model "XWP-70 Insulator", and manufacturing date "2020-05-12"; shooting assistance information may include, for example, shooting time "2024-03-18 10:25:30" and shooting location coordinates "106.78°E, 26.52°N", which can be used to accurately associate the physical location of the transmission line.

[0044] In some exemplary embodiments, a computer device can acquire line images of power transmission lines and image information corresponding to the line images.

[0045] Specifically, computer devices can obtain the line image of the power transmission line and the corresponding image information from the distributed file system corresponding to the target application.

[0046] For example, after the inspection personnel authenticate their identity through the target application, they can enter the "Defect Reporting" function module. First, they can search for and select transmission lines and related equipment in the "Power Grid Equipment" module. The system will automatically synchronize the basic attribute information of the equipment from the equipment information table in the background MySQL database. Then, the inspection personnel can take pictures of the line on-site using the shooting function of the target application. After the pictures are taken, the target application will automatically collect the shooting location coordinates, the shooting timestamp generated by the system time, and the user ID of the currently logged-in inspection personnel as the reporter identifier. Finally, the system will associate and bind the line pictures, synchronized equipment attribute information, and automatically collected shooting auxiliary information to form a complete data group and upload it to the background distributed file system for storage. At the same time, a unique identifier will be generated for subsequent traceability.

[0047] Step 102: Use image information to perform image processing on the line image to obtain the processed line image.

[0048] Optionally, image processing may be one or more combinations of resolution adjustment, image noise reduction, background segmentation, and feature enhancement.

[0049] For example, since line images contain interference factors from the outdoor environment, such as tree obstruction, clouds and fog, excessively strong or dim lighting, and blurring caused by camera shake, and different types of equipment have different requirements for image clarity and detail, environmental interference can be reduced by image processing of line images.

[0050] In some exemplary embodiments, after acquiring the line image of a power transmission line and the image information corresponding to the line image, the computer device can use the image information to perform image processing on the line image to obtain the image-processed line image.

[0051] Specifically, the computer equipment can first determine the quality status and main interference types of the line image based on the shooting auxiliary information in the image information, such as shooting time, ambient brightness correlation data, and equipment correlation information, and then call the preset general image processing algorithm to optimize the line image.

[0052] For example, noise reduction algorithms can be used to remove random noise in outdoor environments, image enhancement techniques can be used to improve the image brightness imbalance caused by uneven lighting, and image stabilization correction can be used to correct image blur caused by camera shake, so as to obtain a processed line image.

[0053] Step 103: Input the processed line image into the defect identification model to obtain the defect identification information of the transmission line output by the defect identification model, and determine the defect information of the transmission line based on the defect identification information.

[0054] Optionally, the defect identification model can be a deep learning model based on an improvement of the YOLO v3 algorithm.

[0055] For example, the defect identification model can be adapted to the defect identification needs of different equipment types in transmission lines. It can automatically capture equipment defect-related features from the processed line images, distinguish normal equipment areas from suspected defect areas, and output preliminary judgment results related to defects.

[0056] Defect information can be used to indicate at least one of the following: the location of a defect, the type of defect, and the level of defect in a transmission line.

[0057] In some exemplary embodiments, after obtaining the image-processed line image, the computer device can input the image-processed line image into the defect identification model to obtain the defect identification information of the transmission line output by the defect identification model.

[0058] Specifically, the computer equipment inputs the processed line image into the defect identification model according to the preset input specifications. The defect identification model uses built-in feature extraction and analysis algorithms to detect suspected defect areas of the transmission line equipment in the processed line image, and outputs the location identifier, feature description, preliminary defect type and level reference, and reliability index corresponding to the judgment result of the area to obtain defect identification information.

[0059] Furthermore, after obtaining the defect identification information of the transmission line output by the defect identification model, the computer equipment can determine the defect information of the transmission line based on the defect identification information.

[0060] Specifically, computer equipment can directly identify defect information as defect information for transmission lines. The computer equipment can also verify the defect information to determine the defect information of the transmission line based on the verification results.

[0061] The aforementioned method for determining defect information of transmission lines first acquires a line image of the transmission line and corresponding image information. The image information indicates the equipment and its attribute information associated with the transmission line image. Then, the image information is used to process the line image to obtain a processed line image. This processed line image is then input into a defect identification model to obtain the defect identification information of the transmission line output by the model. Based on this defect identification information, the defect information of the transmission line is determined. The defect information indicates at least one of the following: defect location, defect type, and defect level. The method for determining defect information of transmission lines provided in this application can determine the defect information of the transmission line based on the line image and corresponding image information, effectively improving the efficiency of defect information determination compared to the prior art's method of determining defect information through manual inspection.

[0062] In one exemplary embodiment, such as Figure 2 As shown, the defect identification model includes a feature extraction unit, a candidate box generation unit, a confidence evaluation unit, and a classification and grading unit. The processed line image is input into the defect identification model to obtain the defect identification information of the transmission line output by the model, including the following steps:

[0063] Step 201: Input the processed line image into the feature extraction unit to obtain the multi-scale feature map output by the feature extraction unit.

[0064] Optionally, the feature extraction unit can be composed of a backbone convolutional layer, a feature selection activation layer, and a multi-scale feature fusion layer, integrating RGB multi-channel perception, Maxout network feature optimization, and multi-scale feature fusion technology.

[0065] For example, since defects in transmission lines vary in size, such as slight wear on conductors, broken insulators, and missing tower components, the visual scale of these defects differs. Furthermore, defect features may be distributed across different layers of the image, such as shallow layers containing detailed features like edges and textures, and deeper layers containing semantic features related to the defect type. Therefore, the feature extraction unit can generate multi-scale feature maps through multi-stage processing to cover defect features of different types and sizes.

[0066] In some exemplary embodiments, a computer device can input an image of a line after image processing into a feature extraction unit to obtain a multi-scale feature map output by the feature extraction unit.

[0067] Step 202: Input the multi-scale feature map into the candidate box generation unit to obtain the candidate boxes output by the candidate box generation unit.

[0068] Optionally, the candidate box generation unit can integrate the k-means clustering algorithm with a preset anchor box matching mechanism, which can accurately locate potential defect areas in feature maps of different scales, avoiding missing small defects or misidentifying non-defect areas.

[0069] Candidate boxes are used to identify candidate defect regions in multi-scale feature maps.

[0070] For example, since the actual size of various defects in transmission lines varies significantly, such as millimeter-level wear on conductor surfaces, overall insulator breakage, and missing tower components, and since different scale feature maps correspond to different size defect features (e.g., shallow feature maps are suitable for small-sized defects, while deep feature maps are suitable for large-sized defects), the candidate box generation unit can accurately match the diverse candidate box sizes obtained through clustering with multi-scale feature maps, ensuring that potential defect areas of different types and sizes can be effectively identified, thus solving the problem of poor adaptability of traditional fixed-size candidate boxes.

[0071] In some exemplary embodiments, a computer device may input multi-scale feature maps into a candidate box generation unit to obtain candidate boxes output by the candidate box generation unit.

[0072] Step 203: Input the candidate bounding box and multi-scale feature map into the confidence evaluation unit to obtain the confidence information of the candidate defect region output by the confidence evaluation unit, and determine the target candidate defect region from the candidate defect region based on the confidence information.

[0073] Optionally, the confidence assessment unit can integrate an intersection-union calculation module, a defect feature matching degree analysis module, and a threshold determination module to evaluate whether the region corresponding to the candidate box is a real defect region through dual dimensions. The confidence information can reflect the accuracy of the candidate box positioning and the reliability of the defect features within the region.

[0074] For example, since the candidate boxes output by the candidate box generation unit may include non-defect areas, such as tree shadows, misidentification of normal equipment structures, or areas with large positioning deviations, and the feature recognition of different types of defects differs, the confidence evaluation unit can quantitatively evaluate each candidate defect area to effectively eliminate false positive candidate boxes, select areas with accurate positioning and clear defect features, and improve the overall defect recognition efficiency and accuracy.

[0075] In some exemplary embodiments, the computer device can input candidate bounding boxes and multi-scale feature maps into a confidence evaluation unit to obtain confidence information of the candidate defect region output by the confidence evaluation unit, and determine the target candidate defect region from the candidate defect region based on the confidence information.

[0076] Specifically, after receiving candidate bounding boxes and multi-scale feature maps, the confidence assessment unit can extract the feature map region corresponding to each candidate bounding box and call a preset defect feature template library. Then, it calculates confidence information through two dimensions: the first dimension is location confidence, which calculates the IOU value between the candidate bounding box and the corresponding defect feature template, i.e., the ratio of the intersection area to the union area of ​​the candidate bounding box region and the template feature region, reflecting the accuracy of the candidate bounding box localization; the second dimension is feature confidence, which calculates the matching similarity between the feature map region corresponding to the candidate bounding box and the defect feature template, such as the cosine similarity of feature vectors, reflecting the degree of fit between the features within the region and the true defect features. Finally, the location confidence and feature confidence are weighted and fused according to preset weights. The comprehensive confidence information of each candidate defect region is obtained, with a value range of [0, 1]. The closer the value is to 1, the higher the reliability. Next, the confidence evaluation unit calls the preset confidence threshold. This threshold is calibrated based on a large amount of historical defect identification data and has preset differentiated thresholds for defects of different equipment types. For example, the threshold for insulator defects is set to 0.7 and the threshold for conductor defects is set to 0.65, which is in line with the differences in the identification difficulty of various defects. Finally, the comprehensive confidence information of each candidate defect region is compared with the corresponding threshold, and candidate defect regions with confidence higher than the threshold are selected. At the same time, non-maximum suppression algorithm is used to remove duplicates of candidate defect regions with overlapping values, and the region with the highest confidence is retained to determine the target candidate defect region.

[0077] Step 204: Input the target candidate defect area into the classification and grading unit to obtain the defect identification information of the transmission line output by the classification and grading unit.

[0078] Optionally, the classification and grading unit can be composed of a defect classification subunit and a defect grading subunit, integrating a logistic regression classification algorithm, a binary cross-entropy loss function, and a power grid defect level mapping mechanism.

[0079] For example, transmission line defects are diverse, such as insulator defects including cracks, contamination, and spontaneous explosions, and conductor defects including wear, broken strands, and foreign object attachments. The severity of different types of defects directly determines the priority of defect elimination. For example, missing tower components are serious defects that require emergency handling, while slight wear on conductors is a general defect that can be handled in a planned manner. Therefore, the classification and grading unit needs to be based on standardized specifications, first clarify the defect type, and then determine the grade by combining the quantitative indicators of defect characteristics to ensure that the output results meet the actual needs of power grid operation and maintenance management.

[0080] In some exemplary embodiments, a computer device may input a target candidate defect region into a classification and grading unit to obtain defect identification information of the transmission line output by the classification and grading unit.

[0081] Specifically, the classification and grading unit can extract deep feature vectors of the target candidate defect region based on the fusion results of semantic and detail features of the region in the multi-scale feature map, and input them into the defect classification subunit. The defect classification subunit uses a logistic regression algorithm to perform multi-label classification processing on the feature vectors, calculates the matching probability between the feature vectors and various defect type templates through a binary cross-entropy loss function, and selects the defect type with the highest matching probability and above the classification threshold as the defect type determination result for the region, while outputting the type determination confidence score. Subsequently, the defect type determination result and the feature quantification indicators of the target candidate defect region, such as defect area ratio, edge damage degree, and feature anomaly intensity, are input into the system. The defect classification subunit calls a preset defect type-classification rule mapping table, calculates a level matching score based on feature quantification indicators, determines the corresponding defect level based on the score, and outputs the level judgment confidence level. The defect type-classification rule mapping table is constructed strictly according to the defect classification standard. For example, insulator rupture defects are classified into levels one to five according to the proportion of rupture area, and conductor strand breakage defects are classified into general, severe, and critical levels according to the number of broken strands. Finally, the classification and classification unit integrates the defect type judgment result, type confidence level, defect level judgment result, level confidence level, and the coordinate position information and feature description information of the target candidate defect area in the original line image to obtain defect identification information.

[0082] In one exemplary embodiment, such as Figure 3 As shown, the feature extraction unit includes a backbone convolutional layer, a feature selection activation layer, and a multi-scale feature fusion layer. The processed line image is input into the feature extraction unit to obtain the multi-scale feature map output by the feature extraction unit, including the following steps:

[0083] Step 301: Input the processed line image into the backbone convolutional layer for convolution and downsampling to obtain the initial first-scale feature map, the initial second-scale feature map, and the initial third-scale feature map.

[0084] Optionally, the backbone convolutional layer can adopt a stacked structure of 1×1 convolutional dimensionality reduction and 3×3 convolutional feature enhancement, integrating weight sharing and local perception mechanisms to adapt to the complex scene features of power transmission line images. This can gradually extract multi-dimensional features from shallow details to deep semantics while reducing computational load.

[0085] For example, since the defect features in transmission line images are distributed in a hierarchical manner, such as the edge texture of worn conductors belonging to shallow detail features, the overall shape of broken insulators belonging to medium-level structural features, and the semantic association of missing tower components belonging to deep features, and different levels of features correspond to different dimensions of defect identification criteria, through multi-stage convolution and downsampling processing of the backbone convolutional layer, redundant information in the image can be gradually stripped away and effective defect features can be enhanced. At the same time, downsampling expands the receptive field, realizing feature capture from local details to global semantics.

[0086] In some exemplary embodiments, a computer device may input the image-processed line image into a backbone convolutional layer for convolution and downsampling to obtain an initial first-scale feature map, an initial second-scale feature map, and an initial third-scale feature map.

[0087] Specifically, the line image can be input into the backbone convolutional layer in RGB three-channel format. The backbone convolutional layer consists of multiple cascaded convolutional modules and downsampling modules, each containing a 1×1 convolutional layer and a 3×3 convolutional layer. The 1×1 convolutional layer is responsible for feature dimension reduction, reducing subsequent computation and preserving core features. The 3×3 convolutional layer extracts local image features based on a local perception mechanism, such as device surface texture and component edge contours. The convolution process uses a weight sharing mechanism to reduce model complexity and improve feature extraction efficiency. Subsequently, three initial feature maps at different scales are generated through three ordered convolution stacks and downsampling operations: First, the processed line image undergoes feature extraction through eight convolutional modules (including alternating 1×1 and 3×3 convolutional layers), and then max pooling downsampling is used to reduce the image size and expand the receptive field to obtain the initial first-scale features. The first step involves inputting the initial first-scale feature map into six subsequent convolutional modules for further feature enhancement, followed by max pooling downsampling to obtain the initial second-scale feature map. This second-scale feature map has been stripped of some redundant background information and focuses on presenting the structural outline and morphological features of the equipment components, supporting feature extraction for medium-sized defects. The third step involves inputting the initial second-scale feature map into the last six convolutional modules for deep feature mining, followed by max pooling downsampling again to obtain the initial third-scale feature map. This third-scale feature map has a receptive field covering the entire image and focuses on deep semantic features, supporting feature capture for large-sized defects. This process yields the initial first-scale feature map, the initial second-scale feature map, and the initial third-scale feature map.

[0088] Step 302: Input the initial first-scale feature map, the initial second-scale feature map, and the initial third-scale feature map into the corresponding feature filtering activation layer for feature filtering and enhancement processing to obtain the target first-scale feature map, the target second-scale feature map, and the target third-scale feature map.

[0089] Optionally, each scale feature map corresponds to an independent feature filtering activation layer. Each layer adopts a Maxout network structure and a dynamic feature weight adjustment mechanism to selectively filter out effective features related to defects in each scale feature map based on the feature distribution pattern of power grid transmission line defects, suppress redundant features such as background interference and normal equipment structure, and enhance the identification of defect features.

[0090] For example, although the initial first-scale feature map contains shallow details such as device edges and textures, it may contain interfering features such as tree shadows and dust noise in the outdoor environment; in the device structure outline of the initial second-scale feature map, normal component features and defect features are easily confused; in the deep semantic features of the initial third-scale feature map, the response strength of defect-related features may be insufficient. Therefore, by using targeted processing of the corresponding feature filtering activation layer, invalid information in the feature maps of each scale can be removed.

[0091] In some exemplary embodiments, the computer device may input the initial first-scale feature map, the initial second-scale feature map, and the initial third-scale feature map into the corresponding feature filtering activation layer for feature filtering and enhancement processing, so as to obtain the target first-scale feature map, the target second-scale feature map, and the target third-scale feature map.

[0092] Specifically, the initial first-scale feature map, initial second-scale feature map, and initial third-scale feature map can be input into the corresponding feature filtering activation layer. Each activation layer first performs RGB multi-channel feature splitting on the input feature map and captures the local feature response within each channel based on the local perception mechanism. Then, the Maxout network structure performs maximum pooling on the split multi-channel features, that is, it takes the maximum value of the multi-channel feature value corresponding to each feature point to filter out the feature components that contribute the most to defect identification and remove redundant and highly interfering feature components, thereby simplifying the feature dimensions and filtering effective features. Next, each activation layer calls the preset scale-defect feature matching template library in the background, calculates the matching degree between the filtered features and the template library, and dynamically adjusts the feature weights according to the matching degree. High weights are given to defect features with high matching degree to further enhance their response strength, while low weights are given to interfering features with low matching degree to weaken their influence. Finally, the target first-scale feature map, target second-scale feature map, and target third-scale feature map are obtained.

[0093] Step 303: Input the target first-scale feature map, target second-scale feature map and target third-scale feature map into the multi-scale feature fusion layer for upsampling and feature stitching to obtain multi-scale feature maps.

[0094] Optionally, the multi-scale feature fusion layer can be composed of an upsampling module and a feature concatenation module.

[0095] For example, while the first-scale feature map of the target retains the details of the defect, it lacks semantic information. While the third-scale feature map of the target has semantic judgment capabilities, it loses details. If a feature map of a certain scale is used alone, it is easy to miss small-sized defects or misjudge large-sized defects. Therefore, the multi-scale feature fusion layer integrates information by upsampling to unify the resolution and by splicing features, so that the final output multi-scale feature map contains both detailed texture and semantic support, which can fully adapt to the recognition needs of different types and sizes of defects.

[0096] In some exemplary embodiments, a computer device can input a first-scale feature map, a second-scale feature map, and a third-scale feature map of the target into a multi-scale feature fusion layer for upsampling and feature stitching to obtain a multi-scale feature map.

[0097] Specifically, the multi-scale feature fusion layer first upsamples the target third-scale feature map by using bilinear interpolation to double its resolution, ensuring its size matches the target second-scale feature map, resulting in an upsampled third-scale feature map. Then, the upsampled third-scale feature map and the target second-scale feature map are input into the feature concatenation module, where a concat operation is performed along the channel dimension to generate the first-stage fusion feature map. Next, the first-stage fusion feature map is upsampled again, with bilinear interpolation to double its resolution, matching its size to the target first-scale feature map, resulting in a second-stage upsampled fusion feature map. Finally, the second-stage upsampled fusion feature map is concatenated along the channel dimension with the target first-scale feature map, integrating the shallow detail features of the target first-scale feature map with the previously fused mid-level structural features and deep semantic features to obtain a multi-scale feature map.

[0098] In one exemplary embodiment, such as Figure 4 As shown, the process of inputting multi-scale feature maps into a candidate box generation unit to obtain candidate boxes output by the candidate box generation unit includes the following steps:

[0099] Step 401: Obtain defect size sample data of transmission lines, and perform cluster analysis on the defect size sample data based on clustering algorithm to obtain a set of candidate box sizes.

[0100] Optionally, the defect size sample data can be historical defect image data verified by manual annotation, and can include common defects in various power transmission equipment such as insulators, conductors, towers, and fittings. The clustering algorithm can be the k-means clustering algorithm.

[0101] In some exemplary embodiments, a computer device can acquire defect size sample data of a power transmission line and perform cluster analysis on the defect size sample data based on a clustering algorithm to obtain a set of candidate box sizes.

[0102] Specifically, the computer equipment can retrieve a preset number of valid defect images that have passed verification and annotation from a defect image library associated with a distributed file system. These images cover various types of equipment, different defect types, and different degrees of severity in power transmission lines, ensuring the comprehensiveness and representativeness of the sample data. Subsequently, the manually annotated defect region bounding box information is extracted from each defect image, and the width and height dimensions of the bounding boxes are parsed to obtain the data. This data is then classified and annotated according to the corresponding equipment type and defect type to form an original defect size sample dataset. Next, the original dataset is preprocessed: outliers are removed and normalization is performed to obtain a standardized defect size sample dataset. Then, the optimal number of clusters k for the k-means clustering algorithm is determined based on the elbow rule. The standardized dataset is then input into the k-means clustering algorithm, and through iterative calculation, the defect size similarity within the same cluster is maximized, and the similarity between different clusters is minimized, ultimately obtaining k cluster centers. The width and height dimensions corresponding to each cluster center are the core dimensions of the candidate boxes. Finally, the core dimensions obtained from clustering are inversely normalized to the original image pixel units to obtain a set of candidate box dimensions.

[0103] Step 402: Determine candidate boxes based on the candidate box size set and multi-scale feature maps.

[0104] In some exemplary embodiments, after obtaining a set of candidate box sizes, the computer device can determine candidate boxes based on the set of candidate box sizes and multi-scale feature maps.

[0105] Specifically, the computer equipment can first classify the candidate bounding box size set, splitting it into three size subsets based on size specifications: a small-sized candidate bounding box subset, a medium-sized candidate bounding box subset, and a large-sized candidate bounding box subset. The aspect ratios of the three subsets all cover common defect shape ratios such as 1:1, 1:2, and 2:1. Then, a scale-size subset mapping relationship is established: the small-sized candidate bounding box subset is assigned to the target first-scale feature map, the medium-sized candidate bounding box subset to the target second-scale feature map, and the large-sized candidate bounding box subset to the target third-scale feature map, ensuring that the candidate bounding box size of each scale feature map matches the corresponding defect size. Next, for each scale feature map, all its feature points are traversed, and a preset feature response evaluation algorithm is called to calculate the relationship between each feature point and its corresponding scale. The matching degree of each candidate box within the subset is used to calculate the cosine similarity with the defect feature template corresponding to the candidate box. The higher the similarity, the greater the probability that the region corresponding to the feature point is a defect. Subsequently, a feature response threshold is set. The thresholds for different scale feature maps are calibrated with historical samples, such as 0.6 for shallow feature maps, 0.55 for medium-scale feature maps, and 0.5 for deep feature maps, to adapt to the differences in response intensity of features at different scales. Feature points with response values ​​higher than the threshold are selected as potential defect centers. Finally, based on each potential defect center, and combined with the sizes of all candidate boxes within the corresponding size subset, multiple rectangular candidate boxes containing the feature point are generated. At the same time, the coordinate position of each candidate box in the feature map of the corresponding scale is recorded, and its absolute coordinates in the original line image are converted through a coordinate mapping algorithm to obtain the candidate box.

[0106] In one exemplary embodiment, such as Figure 5 As shown, image processing is performed on the line image using image information to obtain the processed line image, including the following steps:

[0107] Step 501: Determine the equipment type of the equipment in the associated transmission line based on the image information.

[0108] For example, transmission lines can include various equipment types such as insulators, 110kV conductors, tension towers, fittings, and grounding devices. The structural forms, defect characteristics, and image processing requirements of different equipment vary significantly. For instance, insulator defects are mostly surface cracks and dirt, requiring enhanced detail contrast; conductor defects are mostly wear and broken strands, requiring highlighted edge features. If uniform image processing parameters are used without clearly defining the equipment type, it will lead to insufficient enhancement of defect features or amplification of interference features. Therefore, the equipment type can be determined through image information to provide clear parameter basis for resolution adjustment, feature enhancement, and other processing, thereby improving the effectiveness of image processing.

[0109] In some exemplary embodiments, a computer device may determine the device type of a device in an associated transmission line based on image information.

[0110] Specifically, the computer equipment extracts relevant fields related to the equipment type from the image information. These fields include equipment identification information automatically synchronized by the system when inspectors search for and select equipment in the defect reporting module of the target application, as well as equipment type codes retrieved from the equipment information table in the MySQL database. Subsequently, the computer equipment matches the extracted equipment type codes with the standard equipment type codes in the parameter table. The table stores the standardized codes and name mappings for all equipment types of transmission lines. By matching the codes, the equipment type name can be initially determined. Next, to ensure the accuracy of the determination, the system further performs cross-validation by combining auxiliary fields such as equipment model and voltage level of the line in the image information to eliminate possible deviations in code matching. Finally, if all relevant fields point to the same equipment type, the standardized equipment type result is output. If field conflicts or missing information occur, the system will trigger an exception handling mechanism, calling the target application to pop up a window to remind inspectors to supplement and confirm the equipment type, or automatically supplementing the information based on the coordinates of the shooting location and the power grid equipment ledger.

[0111] Step 502: Adjust the resolution of the line image according to the equipment type to obtain the processed line image.

[0112] For example, the defect feature scales of different equipment types vary significantly: defects in insulators are mostly small surface cracks and dirt spots, which require high-resolution images to be clearly captured; defects in 110kV conductors, such as broken strands and foreign objects attached, require medium resolution to balance detail and computational efficiency; defects in tension towers, such as missing parts and structural deformation, do not require excessively high resolution.

[0113] In some exemplary embodiments, after determining the device type of the equipment in the associated transmission line based on image information, the computer device can adjust the resolution of the line image according to the device type to obtain a processed line image.

[0114] Specifically, the computer device can access a preset device type-optimal resolution mapping table, which is stored in a parameter table of a MySQL database. The resolution parameters of this table are calibrated using a large number of defect samples, ensuring both the input size compatibility of the improved YOLO v3 model and the defect identification requirements of various devices. Subsequently, the original resolution information of the line image is extracted and compared with the optimal resolution corresponding to the current device type in the mapping table. If the original resolution is higher than the optimal resolution, a bilinear interpolation algorithm is used for proportional scaling down, preserving the integrity of defect features while reducing resolution and avoiding image blurring. If the original resolution is lower than the optimal resolution, a bicubic interpolation algorithm is used for proportional scaling up, supplementing image details and ensuring that defect features are not compressed or distorted. If the original resolution matches the optimal resolution, the original image resolution is directly retained. Finally, the adjusted image undergoes resolution verification to check whether its aspect ratio and pixel size meet preset standards. If deviations exist, a second fine-tuning is performed to obtain the processed line image.

[0115] In one exemplary embodiment, such as Figure 6 As shown, determining the defect information of a transmission line based on defect identification information includes the following steps:

[0116] Step 601: Determine whether the defect identification information meets the preset verification conditions.

[0117] Optionally, the preset verification conditions can be pre-set by technicians according to actual needs.

[0118] For example, if the defect identification information is "insulator rupture, type confidence 0.68, level confidence 0.62, rupture area percentage 15%", the corresponding preset verification conditions can be: insulator type defect confidence ≥ 0.7, level confidence ≥ 0.6, rupture area percentage conforming to the quantitative range of level 2 defects in the standard library (10%-20%), and "rupture" type logically matching level 2. In this case, the identification information does not meet the verification conditions because the type confidence does not meet the standard. If the identification information is "conductor strand breakage, type confidence 0.75, level confidence 0.71, number of broken strands 2", the corresponding conductor defect verification conditions can be: type confidence ≥ 0.65, level confidence ≥ 0.6, and the number of broken strands matching the severity level requirement. In this case, the verification conditions are met.

[0119] In some exemplary embodiments, the computer device can determine whether the defect identification information meets preset verification conditions.

[0120] Specifically, the computer equipment can extract core verification parameters from the defect identification information, including defect type, type confidence level, defect level, level confidence level, feature quantification index, and target candidate defect area coordinate information. Subsequently, it calls the equipment type-verification condition mapping relationship stored in the MySQL database parameter table to obtain the preset verification rules corresponding to the current equipment type, including: 1. Quantification thresholds: minimum threshold for type confidence level, minimum threshold for level confidence level, and compliance range for feature quantification index; 2. Qualitative specifications: the defect type must be within the corresponding equipment type defect list in the "Transmission Equipment Defect Standard Library," and the defect level... The criteria must match the feature quantification indicators, and the coordinates of the target candidate defect area must be within the valid range of the original line image. Next, the verification is performed in the order of prior quantification indicators and subsequent qualitative specifications: first, it is determined whether the type confidence and level confidence are higher than the corresponding thresholds, and whether the feature quantification indicators fall within the compliance range; then, it is checked whether the defect type conforms to the standard library specifications, whether the type and level are logically compatible, and whether the coordinates are valid; finally, if all quantification indicators meet the standards and all qualitative specifications are satisfied, the defect identification information is determined to meet the preset verification conditions; if any indicator fails to meet the standards or the specifications are not satisfied, the preset verification conditions are determined not to be satisfied.

[0121] Step 602: If satisfied, determine the defect information of the transmission line based on the defect identification information.

[0122] In some exemplary embodiments, if a computer device determines that the defect identification information meets preset verification conditions, it can determine the defect information of the transmission line based on the defect identification information.

[0123] Specifically, computer equipment can identify defect information as defect information of power transmission lines.

[0124] Step 603: If the conditions are not met, the defect identification information is corrected, and the defect information of the transmission line is determined based on the corrected defect identification information.

[0125] In some exemplary embodiments, if the computer device determines that the defect identification information does not meet the preset verification conditions, it can perform correction processing on the defect identification information and determine the defect information of the transmission line based on the corrected defect identification information.

[0126] Specifically, computer equipment can trigger a manual calibration process, generating a calibration task package containing circuit images, original defect identification information, explanations of non-compliant verification items, and reference materials from the standard library. This package is then assigned to the inspection team leader or expert verification end based on the complexity of the defects. Verifiers, in conjunction with standard library clauses and similar cases, specifically correct the defect type, level, characteristic quantitative indicators, or candidate area coordinates, and submit the calibration results. The system performs a second verification of the corrected information. If it meets the standards, it is directly organized into standardized defect information containing the defect location, type, and level and stored. If it does not meet the standards, it is upgraded and pushed to a higher-level expert for review until the verification conditions are met.

[0127] In one exemplary embodiment, such as Figure 7 As shown, another method for determining defect information of transmission lines is provided, which includes the following steps:

[0128] Step 701: Obtain the line image of the transmission line and the corresponding image information; the image information is used to indicate the equipment in the transmission line associated with the line image and the attribute information of the equipment; determine the equipment type of the equipment in the associated transmission line based on the image information; adjust the resolution of the line image according to the equipment type to obtain the processed line image.

[0129] Step 702: Input the processed line image into the backbone convolutional layer of the feature extraction unit in the defect recognition model for convolution and downsampling to obtain initial first-scale feature maps, initial second-scale feature maps, and initial third-scale feature maps; input the initial first-scale feature maps, initial second-scale feature maps, and initial third-scale feature maps into the feature filtering activation layers of the corresponding feature extraction units for feature filtering and enhancement to obtain target first-scale feature maps, target second-scale feature maps, and target third-scale feature maps; input the target first-scale feature maps, target second-scale feature maps, and target third-scale feature maps into the multi-scale feature fusion layer of the feature extraction unit for upsampling and feature concatenation to obtain multi-scale feature maps;

[0130] Step 703: Obtain defect size sample data of the transmission line, and perform cluster analysis on the defect size sample data based on a clustering algorithm to obtain a set of candidate box sizes; determine candidate boxes based on the candidate box size set and multi-scale feature map; the candidate boxes are used to identify candidate defect regions in the multi-scale feature map; input the candidate boxes and multi-scale feature map into the confidence evaluation unit in the defect identification model to obtain the confidence information of the candidate defect regions output by the confidence evaluation unit, and determine the target candidate defect region from the candidate defect regions based on the confidence information; input the target candidate defect region into the classification and grading unit in the defect identification model to obtain the defect identification information of the transmission line output by the classification and grading unit;

[0131] Step 704: Determine whether the defect identification information meets the preset verification conditions; if it does, determine the defect information of the transmission line based on the defect identification information; if it does not meet the conditions, correct the defect identification information and determine the defect information of the transmission line based on the corrected defect identification information; the defect information is used to indicate at least one of the defect location, defect type and defect level of the transmission line.

[0132] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0133] Based on the same inventive concept, this application also provides a device for determining defect information of transmission lines to implement the aforementioned method for determining defect information of transmission lines. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the device for determining defect information of transmission lines provided below can be found in the limitations of the method for determining defect information of transmission lines described above, and will not be repeated here.

[0134] In one exemplary embodiment, such as Figure 8 As shown, a defect information determination device 800 for transmission lines is provided, comprising: an acquisition module 801, an execution module 802, and a determination module 803, wherein:

[0135] The acquisition module 801 is used to acquire the line image of the transmission line and the image information corresponding to the line image; the image information is used to indicate the equipment in the transmission line associated with the line image and the attribute information of the equipment;

[0136] Execution module 802 is used to perform image processing on the line image using image information to obtain the image-processed line image;

[0137] The determination module 803 is used to input the image of the line after image processing into the defect recognition model to obtain the defect recognition information of the transmission line output by the defect recognition model, and to determine the defect information of the transmission line based on the defect recognition information; the defect information is used to indicate at least one of the defect location, defect type and defect level of the transmission line.

[0138] In one embodiment, the defect identification model includes a feature extraction unit, a candidate box generation unit, a confidence assessment unit, and a classification and grading unit. The determination module 803 is specifically used to input the image-processed line image into the feature extraction unit to obtain a multi-scale feature map output by the feature extraction unit; input the multi-scale feature map into the candidate box generation unit to obtain candidate boxes output by the candidate box generation unit; the candidate boxes are used to identify candidate defect regions in the multi-scale feature map; input the candidate boxes and the multi-scale feature map into the confidence assessment unit to obtain confidence information of the candidate defect regions output by the confidence assessment unit, and determine the target candidate defect region from the candidate defect regions based on the confidence information; input the target candidate defect region into the classification and grading unit to obtain defect identification information of the transmission line output by the classification and grading unit.

[0139] In one embodiment, the feature extraction unit includes a backbone convolutional layer, a feature selection activation layer, and a multi-scale feature fusion layer; the determination module 803 is specifically used to input the image-processed line image into the backbone convolutional layer for convolution and downsampling processing to obtain an initial first-scale feature map, an initial second-scale feature map, and an initial third-scale feature map; input the initial first-scale feature map, the initial second-scale feature map, and the initial third-scale feature map into the corresponding feature selection activation layer for feature selection and enhancement processing to obtain a target first-scale feature map, a target second-scale feature map, and a target third-scale feature map; and input the target first-scale feature map, the target second-scale feature map, and the target third-scale feature map into the multi-scale feature fusion layer for upsampling and feature concatenation processing to obtain a multi-scale feature map.

[0140] In one embodiment, the determining module 803 is specifically used to acquire defect size sample data of the transmission line, and perform cluster analysis on the defect size sample data based on a clustering algorithm to obtain a set of candidate box sizes; and determine candidate boxes based on the set of candidate box sizes and multi-scale feature maps.

[0141] In one embodiment, the execution module 802 is specifically used to determine the equipment type of the equipment in the associated transmission line based on image information; and to adjust the resolution of the line image according to the equipment type to obtain a processed line image.

[0142] In one embodiment, the determining module 803 is specifically used to determine whether the defect identification information meets the preset verification conditions; if it does, the defect information of the transmission line is determined based on the defect identification information; if it does not, the defect identification information is corrected, and the defect information of the transmission line is determined based on the corrected defect identification information.

[0143] Each module in the aforementioned power transmission line defect information determination device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0144] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When executed by the processor, the computer program implements a method for determining defect information in power transmission lines.

[0145] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a method for determining defect information in power transmission lines.

[0146] Those skilled in the art will understand that Figure 9 and Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0147] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the methods described in any of the above embodiments.

[0148] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the methods described in any of the above embodiments.

[0149] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the method described in any of the above embodiments.

[0150] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0151] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0152] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for determining defect information of a power transmission line, characterized in that, The method includes: Obtain a line image of a power transmission line and corresponding image information; the image information is used to indicate the equipment in the power transmission line associated with the line image and the attribute information of the equipment; The image information is used to perform image processing on the line image to obtain a processed line image; The processed line image is input into the defect identification model to obtain the defect identification information of the transmission line output by the defect identification model, and the defect information of the transmission line is determined based on the defect identification information; the defect information is used to indicate at least one of the defect location, defect type and defect level of the transmission line.

2. The method according to claim 1, characterized in that, The defect identification model includes a feature extraction unit, a candidate box generation unit, a confidence evaluation unit, and a classification and grading unit; the step of inputting the processed line image into the defect identification model to obtain the defect identification information of the transmission line output by the defect identification model includes: The processed line image is input into the feature extraction unit to obtain the multi-scale feature map output by the feature extraction unit. The multi-scale feature map is input into the candidate box generation unit to obtain candidate boxes output by the candidate box generation unit; the candidate boxes are used to identify candidate defect regions in the multi-scale feature map. The candidate bounding box and the multi-scale feature map are input to the confidence evaluation unit to obtain the confidence information of the candidate defect region output by the confidence evaluation unit, and the target candidate defect region is determined from the candidate defect region based on the confidence information; The target candidate defect area is input into the classification and grading unit to obtain the defect identification information of the transmission line output by the classification and grading unit.

3. The method according to claim 2, characterized in that, The feature extraction unit includes a backbone convolutional layer, a feature selection activation layer, and a multi-scale feature fusion layer; the step of inputting the processed line image into the feature extraction unit to obtain the multi-scale feature map output by the feature extraction unit includes: The processed line image is input into the backbone convolutional layer for convolution and downsampling to obtain an initial first-scale feature map, an initial second-scale feature map, and an initial third-scale feature map. The initial first-scale feature map, the initial second-scale feature map, and the initial third-scale feature map are respectively input into the corresponding feature filtering activation layer for feature filtering and enhancement processing to obtain the target first-scale feature map, the target second-scale feature map, and the target third-scale feature map; The target's first-scale feature map, second-scale feature map, and third-scale feature map are input into the multi-scale feature fusion layer for upsampling and feature stitching to obtain the multi-scale feature map.

4. The method according to claim 2, characterized in that, The step of inputting the multi-scale feature map into the candidate box generation unit to obtain the candidate boxes output by the candidate box generation unit includes: Obtain defect size sample data of the transmission line, and perform cluster analysis on the defect size sample data based on a clustering algorithm to obtain a set of candidate box sizes; The candidate boxes are determined based on the set of candidate box sizes and the multi-scale feature map.

5. The method according to claim 1, characterized in that, The step of processing the line image using the image information to obtain a processed line image includes: The equipment type of the associated equipment in the transmission line is determined based on the image information; The resolution of the line image is adjusted according to the device type to obtain the processed line image.

6. The method according to any one of claims 1 to 5, characterized in that, Determining the defect information of the transmission line based on the defect identification information includes: Determine whether the defect identification information meets the preset verification conditions; If the conditions are met, the defect information of the transmission line is determined based on the defect identification information. If the conditions are not met, the defect identification information is corrected, and the defect information of the transmission line is determined based on the corrected defect identification information.

7. A device for determining defect information of a power transmission line, characterized in that, The device includes: The acquisition module is used to acquire a line image of a power transmission line and image information corresponding to the line image; the image information is used to indicate the equipment in the power transmission line associated with the line image and the attribute information of the equipment; The execution module is used to perform image processing on the line image using the image information to obtain the image-processed line image; The determination module is used to input the image of the line after image processing into the defect recognition model to obtain the defect recognition information of the transmission line output by the defect recognition model, and to determine the defect information of the transmission line based on the defect recognition information; the defect information is used to indicate at least one of the defect location, defect type and defect level of the transmission line.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images